Sphinx - C#: leaderboard
Sphinx checklist-coverage score on its 500 C# pull requests, 450 needing changes and 50 bug-free controls whose correct review is no comment: a GPT-4o judge checks each generated review against a checklist derived from the merged change, buggy cases weighted 0.9; percent, higher is better. Microsoft and University of Rochester, 2026.
Metric: Checklist Coverage Score (%; GPT-4o judge, 450 buggy + 50 bug-free PRs per language). Source: arxiv.org. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini | 40.01 | #180 |
| 2 | GPT-4.1 | 38.62 | #248 |
| 3 | Claude 3.7 Sonnet | 37.68 | #249 |
| 4 | Claude 3.5 Sonnet | 37.16 | #350 |
| 5 | Gemini 2.5 Pro | 36.75 | #152 |
| 6 | Qwen 2.5 72B Instruct | 35.06 | #454 |
| 7 | O3 Mini | 34.94 | #272 |
| 8 | GPT-4o (2024-08-06) | 34.32 | #336 |
| 9 | Qwen 3 235B A22B | 34.31 | #313 |
| 10 | GPT-4o Mini | 34.13 | #599 |
| 11 | Qwen 3 30B A3B | 33.03 | #493 |
| 12 | GPT-4.1 Mini | 32.38 | #355 |
| 13 | GPT-4.1 Nano | 32.14 | #731 |
| 14 | DeepSeek V3 | 31.39 | #321 |
| 15 | DeepSeek R1 | 30.09 | #251 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sphinx-c-sharp · How It Works · Data refreshed daily, snapshot 2026-10-20.